The process of language acquisition has long been a point of uncertainty in research exploring the roots of human language. Researchers have conducted extensive investigations to understand the initial state and process of language acquisition, providing insights into how environmental and genetic factors interact to fashion language and cognitive function, and the mechanisms underlying brain plasticity (Weaver et al., 2004; Werker and Tees, 2005; Barkat et al., 2011; Werker and Hensch, 2015). It is now widely accepted that both genetic and experiential factors contribute to language acquisition (Werker and Curtin, 2005; Gervain and Mehler, 2010), and researchers are interested in understanding how these factors interact during human development.

Infants at birth already exhibit advanced speech perception and language learning abilities. Newborns manifest a preference for speech over non-speech sounds (Vouloumanos and Werker, 2007), can discriminate between different languages based on their speech rhythms (Ramus et al., 2000), detect word boundaries (Christophe et al., 2001), discriminate words with different patterns of stress (Sansavini et al., 1997), or even distinguish consonant sounds (Cabrera and Gervain, 2020) and encode voice pitch in an adult-like manner (Arenillas-Alcón et al., 2021). These findings support the role of a genetically driven cerebral organization towards processing specific speech characteristics.

However, the prenatal period is not devoid of language experience and the study of its influence on the newborn’s speech and language encoding capacities is receiving increasing attention. Hearing becomes functional and undergoes most of its development around the 26th to 28th week of gestation, allowing the fetus to perceive the maternal speech signal (Ruben, 1995; Moore and Linthicum, 2007; Granier-Deferre et al., 2011; May et al., 2011; Anbuhl et al., 2016). Although the exact characteristics of the acoustic signal reaching the fetus are not fully understood, intrauterine recordings from animal models and simulations suggest that the maternal womb acts as a low-pass filter, attenuating around 30 dB for frequencies over 600–1,000 Hz (Gerhardt and Abrams, 2000). The low-frequency components of speech that are transmitted through the uterus include pitch, slow aspects of rhythm and some phonetic information (Moon and Fifer, 2000; May et al., 2011). Evidence indicates that prenatal exposure to speech, despite attenuated by the filtering properties of the womb, shapes speech perception and linguistic preferences of newborns, as shown by studies revealing that neonates can recognize a story heard frequently in utero (DeCasper and Spence, 1986), prefer the voice of their mother (DeCasper and Fifer, 1980) and prefer their native language (Moon et al., 1993). Additionally, prenatal learning extends beyond these common preferences. Recent findings indicate that infants acquire specific knowledge of the prosody (Gervain, 2018) and prefer the rhythmic patterns of the language they were exposed to while in utero (Mariani et al., 2023), indicating a very early specialization for their native language.

Yet, with reported rates of bilingualism of around 65% in Europe (Luk, 2017), an open question remains on the influence of prenatal exposure to more than one language on neural plasticity. Over the past 20 years, mounting evidence has suggested that both exposure to a bilingual acoustic environment and learning several languages affects not only language acquisition but a wide range of developmental processes including perception, cognition and brain development (Byers-Heinlein et al., 2019). Prior research has highlighted that early exposure to language influences infants’ acquisition of speech sounds, indicating that, at birth, infants are able to discriminate all phonetic contrasts. As infants age, their perceptual systems are tuned to collapse over phonetic contrasts not found in the input language or languages, such that their ability to distinguish between phonetic elements becomes increasingly specific to their native language(s) (Kuhl et al., 2006; Saffran et al., 2006; Gervain and Werker, 2008; Kovács and Mehler, 2009; Bosch and Sebastián-Gallés, 2010). Moreover, cross-language interactions modulate almost every level of language processing, including speech perception, phonological, vocabulary and semantic development [for comprehensive review, refer to Hammer et al. (2014) and Kroll et al. (2012)]. Furthermore, some bilinguals switch from one language to the other within the same sentence, demonstrating greater demands on cognitive control than monolinguals to navigate the potential cross-language competition considering that language production is equivalent (Kovács and Mehler, 2009).

Speaking two languages daily also has consequences for the way in which higher cognitive processes operate and results in more precocious development of inhibition and attentional abilities (Costa et al., 2008; Kovács and Mehler, 2009; for review see Barac et al., 2014; Bialystok, 2017). There is evidence for functional and structural brain changes associated with bilingualism, even after brief periods of second-language learning (for extensive review see Li et al., 2014). Bilingual infants show different brain responses to native and non-native speech sounds than monolingual infants (Conboy and Kuhl, 2011). Bilingualism also affects the structure of both grey (Ressel et al., 2012) and white matter (Kuhl et al., 2016) in adults. The observed advantages in cognitive control and attentional abilities, as well as the pattern of structural differences, are modulated by the age of second language acquisition, whether the two languages were acquired simultaneously from birth or sequentially later in life and the interaction between languages (Kroll et al., 2012; Barac et al., 2014; Li et al., 2014).

As bilingual mothers speak using two different sets of phonemic categories and even use two slightly different voice pitch ranges (e.g., Ordin and Mennen, 2017), in-utero bilingual environments are characterized by a greater complexity of the reaching speech signal than monolingual ones. Interestingly, neonates exposed prenatally to a bilingual environment can discriminate their two native languages already at birth and exhibit equal preferences for both (Byers-Heinlein et al., 2010). Thus, it appears clear that linguistic experiences while in utero play a significant role in shaping the early development of speech processing. However, how different prenatal maternal linguistic exposure influences the neural mechanisms underlying speech sound processing at birth is currently unknown.

A large body of evidence has supported the study of the neural encoding of speech sounds through electrophysiological recordings. In particular, the frequency-following response (FFR) can provide insights into the underlying neural mechanisms associated with prenatal language experience, shedding light on how early linguistic exposure shapes the speech-encoding capacities of newborns. The FFR is an auditory evoked potential elicited by periodic complex sounds that reflects neural synchronization with the auditory eliciting signal along the ascending auditory pathway (Skoe and Kraus, 2010; Krizman and Kraus, 2019), providing an accurate snapshot of the neural encoding of speech sounds. FFR recordings have thus become a useful tool to investigate the ability to distinguish between the pitch of different speakers’ voices and the ability to encode the fine spectrotemporal details that distinguish different voiced speech sounds (Gorina-Careta et al., 2022). The interest in the neonatal FFR arises from its potential to serve as a predictive measure for future language development (Schochat et al., 2017), since alterations in FFR patterns in children have been associated with difficulties in reading and learning, dyslexia, impairments in phonological awareness and even autism (King et al., 2002; Banai et al., 2009; Chandrasekaran et al., 2009; Basu et al., 2010; Hornickel et al., 2012; Lam et al., 2017; Otto-Meyer et al., 2018; Font-Alaminos et al., 2020; Rosenthal, 2020). Interestingly, the FFR reflects the impact of a wide range of auditory experiences in children and adults, including training interventions, musical practice and bilingualism (Russo et al., 2005; Song et al., 2008; Kraus and Chandrasekaran, 2010; Krizman et al., 2012, 2015; Carcagno and Plack, 2017; Skoe et al., 2017; Gorina-Careta et al., 2019). In adults it has been observed that bilingual experience enhances the neural responses to the fundamental frequency of sounds (Krizman et al., 2015; Skoe et al., 2017), as well as the subcortical representation of pitch-relevant information (Krizman et al., 2012) and neural consistency, which correlated with both a better attentional control and language proficiency (Krizman et al., 2014). In neonates, FFR recordings have also been used to study the effects at birth of prenatal fetal auditory experiences such as music exposure (Arenillas-Alcón et al., 2023), but the influence of prenatal maternal bilingual speech remains unexplored.

In the present study, we aimed to examine the influence of maternal bilingual linguistic exposure in-utero in speech sound encoding at birth. To that end, we recorded FFRs from newborns who had been exposed to either a monolingual or a bilingual fetal environment during the last trimester of gestation and analyzed their capacity to encode voice pitch and vocalic formant structure information.

Frequency following responses (FFR) elicited by a two-vowel speech stimulus /oa/ (Figure 1A) were collected from a total sample of 129 newborns divided into two groups according to their prenatal fetal exposure to monolingual (MON) or bilingual (BIL) maternal speech. To comprehensively evaluate the neonates’ ability to encode the pitch and vowel formant structure of speech sounds, the neural responses to the fundamental frequency (F₀) and the vowels’ first formant (F₁) were analyzed considering the distinct sound characteristics of the different stimulus sections. All detailed descriptive statistics from the parameters analyzed can be found in Supplementary Table S1.

The present study investigated the impact of maternal bilingual speech during pregnancy on the neural encoding of speech pitch and vowel formant structure in neonates. A total sample of 129 healthy-term newborns was divided into two groups according to their monolingual or bilingual prenatal exposure during the last trimester of gestation, as reported by their mothers through a questionnaire. FFRs elicited to a two-vowel speech stimulus /oa/ (Arenillas-Alcón et al., 2021) were recorded to assess the neural responses to the stimulus’ fundamental frequency (F₀ = 113 Hz; related to voice pitch encoding) and the first formant of each vowel (/o/ F₁ = 452 Hz; /a/ F₁ = 678 Hz; related to vowel formant structure encoding). Our results revealed that the neural representation of pitch, as indexed by the spectral amplitude of the FFR_ENV at the stimulus F₀, did not differ between monolingual and bilingual exposure groups, but monolingually exposed neonates exhibited a higher signal-to-noise ratio (SNR) at the F₀ spectral peak, suggesting the contribution of a higher spectral noise at neighboring frequencies in the bilingual group. Additionally, monolingually exposed neonates exhibited larger spectral amplitudes and SNRs of the FFR_TFS at the formant peak frequencies (F₁) of the speech stimulus used, indicating a stronger encoding of vocalic structure. Furthermore, no significant group differences were observed in neural lag and pre-stimulus root mean square (RMS) amplitude, implying comparable neural transmission delays and absence of a distinct overall neural activity prior to the auditory stimulation. Together, these findings provide novel insights into the effects of prenatal language exposure on the neural encoding of speech sounds at birth.

Pitch is a crucial attribute in the perception of periodic speech sounds, as it conveys prosodic information, facilitates speaker recognition and speech segmentation, accelerates phoneme acquisition in tonal languages, helps with language comprehension in noisy environments and even contributes to the perception of the emotional state in a conversation (Musacchia et al., 2007; Benavides-Varela et al., 2012; Partanen et al., 2013a; Plack et al., 2014; Gervain, 2018; Cabrera and Gervain, 2020; Arenillas-Alcón et al., 2021; Ribas-Prats et al., 2021). The fact that neural mechanisms underlying voice pitch encoding are already mature at birth (Jeng et al., 2011; Ribas-Prats et al., 2019; Cabrera and Gervain, 2020; Arenillas-Alcón et al., 2021) suggests that pitch may play a crucial role in the very first stages of language acquisition (Jeng et al., 2016). Going a step further, pitch could provide a neural synchrony channel onto which separate neural representations of other speech features would anchor as parts of an ensemble that would, ultimately, give rise to a coherent percept (Eggermont, 2001).

Previous studies demonstrated that pitch and pitch contour discrimination drastically improve with training (e.g., Carcagno and Plack, 2017). In this regard, growing up in a bilingual environment, which is characterized as more demanding, dynamic, phonologically rich and requiring heightened attention to all linguistic input, is related to a strengthened neural representation of pitch (Krizman et al., 2012, 2015). Different languages have distinct overall height pitch levels. For example, Catalan was observed to have a higher pitch compared to Spanish (Marquina Zazura, 2011); Polish was found to have a higher pitch compared to American English (Majewski et al., 1972); Mandarin, a higher pitch than English (Keating and Kuo, 2012); Japanese, a higher pitch than Dutch (Van Bezooijen, 1995); or Slavic languages, a higher pitch than Germanic ones (Andreeva et al., 2014). Further, speakers of two phonologically similar dialects exhibit differences in their height pitch levels (e.g., two different dialects of Mandarin; Deutsch et al., 2009).

Yet, pitch height is not the only element that contributes significantly to the distinctiveness of a particular language. The intonational patterns, which are the rising and falling patterns of pitch that convey meaning and contribute to the rhythm of speech, may differ between the different languages. When a speaker switches between languages they naturally adjust the specific contours, pitch ranges, and other prosodic features to conform to the norms of the target language, and many linguistic features such as intonation, may affect the mean fundamental frequency of speech (Järvinen et al., 2013). This adjustment helps maintaining communicative clarity and aligns with the phonetic characteristics of the language being spoken (Mary and Yegnanarayana, 2008; Passoni et al., 2022).

With continued exposure to these complex linguistic contexts, the auditory system gradually becomes finely tuned to process sound more efficiently (Krizman et al., 2012). Thus, individuals with years of exposure and interaction with bilingual environments develop enhanced flexibility and speech-encoding abilities. Most notably, previous studies have shown that bilingual individuals, particularly females, exhibit different pitch frequency ranges depending on the language they speak (Ordin and Mennen, 2017). As both pitch and the intonational patterns of the languages are different, and the prosodic elements of speech which include pitch contours, rhythm, and stress (Moon and Fifer, 2000) are acoustic features reliably transmitted through the womb (Gerhardt and Abrams, 2000; May et al., 2011), bilingual mothers provide their children with a higher pitch variability in utero.

Considering the reviewed literature, if the developing auditory system of a fetus, who underwent approximately 3 months of noninteractional exposure to degraded speech, responded to acoustic exposure as the mature one, we would expect newborns from bilingual mothers to exhibit a higher neural encoding of voice pitch. But our results showed otherwise. We found no differences across groups in FFR_ENV spectral amplitudes at F₀, which aligns with the idea that pitch processing mechanisms are already mature at birth. Yet, we observed a decreased SNR at the F₀ in newborns who were prenatally exposed to a bilingual environment. We attempt to reconcile our seemingly contradicting results by hypothesizing that the higher spectral amplitudes found in bilingually exposed neonates at F₀ neighboring frequencies reflect an increased sensitivity to a wider range of pitch frequencies without yet generating a particularly strong response at any of them.

This view aligns with research on perceptual phonetic development, especially when growing in bilingual environments. Previous studies demonstrated that experience with language shapes infants’ abilities to process speech sounds and, with age, the newborn’s ability to differentiate phonetic distinctions becomes more language-specific (Kuhl et al., 2006; Saffran et al., 2006; Gervain and Werker, 2008; Bosch and Sebastián-Gallés, 2010). At birth all infants possess the ability to perceive all sound distinctions used in languages as they are sensitive to the basic rhythmic differences between languages (Nazzi et al., 1998; Byers-Heinlein et al., 2010). Around 3–4 months of age infants are sensitive to rhythmic differences between languages that go beyond their belonging to the three basic rhythmic classes (Bosch and Sebastián-Gallés, 2010; Molnar et al., 2014) and by the age of 6 months monolingual infants’ ability to perceive speech becomes tailored to their native language. Infants exposed to two languages are also able to discriminate the sound contrasts of both their languages, but this occurs only at the end of their first year (Bosch and Sebastián-Gallés, 2003; Sundara et al., 2008; for review see Hammer et al., 2014).

Yet, the early prenatal impact of language goes beyond language discrimination. As reviewed in the introduction, newborns prefer their mother’s voice over other female voices (DeCasper and Fifer, 1980), their communicative cries reflect the prosody of the language they heard in utero (Mampe et al., 2009) and can recognize stories heard during pregnancy (DeCasper and Spence, 1986). Moreover, previous studies also demonstrated that differences in prenatal language exposure modulate perceptual grouping biases at birth (Abboub et al., 2016) and suggest that hearing pitch contrasts before birth may influence pitch-based grouping preferences and may lead to a stable bias at birth. Thus, despite the discrimination (or no discrimination) of languages at birth, prenatal language exposure modulates the processing of speech sounds. Our findings align with the suggested hypothesis that being bilingual confers a greater perceptual flexibility (Abboub et al., 2016), as we observed in bilingually exposed newborns an increased sensitivity to a wider range of pitch frequencies.

Our results also reveal a modulation of the neural encoding of vowel formants (F₁) depending on prenatal linguistic exposure. In particular, monolingual-exposed neonates exhibited higher spectral amplitudes at the corresponding formant frequencies of the stimulus’ /o/ and steady−/a/ vowels. In a previous study, we found that while the neural encoding of pitch was adult-like at birth, formant encoding was still immature (Arenillas-Alcón et al., 2021). As vowel formant center frequencies are language specific and stable regardless of voice pitch variation, which also presents slight modulations in monolingual individuals during natural speaking, the auditory system of a monolingual-exposed fetus receives a more consistent phonetic repertoire than that of a bilingual-exposed. This would possibly lead to a more effective and accurate encoding of the specific language vowel sound characteristics at birth. Simply put, monolingual newborns seem to have an advantage in processing the specific sounds of their mother tongue, a finding previously attributed to postnatal linguistic exposure (Kuhl, 2010). Our findings thus highlight the greater variability of acoustic speech inputs to which the fetus of bilingual mothers would be exposed and therefore suggest the need for bilinguals to develop a different phonological representation for each of the languages (Sebastian-Gallés et al., 2006). Further investigation into the developmental trajectories of auditory processing in different populations of newborns, with different prenatal auditory experiences, and using language-specific phonetic contrasts (e.g., Catalan contrasts such as /e - ɛ/), which are especially difficult –when not impossible– to detect for Spanish-monolinguals (Pallier et al., 1997, 2001), may shed more light on this issue.

Despite being confident about our results due to the abovementioned reasons, we are fully aware of a number of limitations of our study: language exposure was assessed by a short (approx. 5 min answer time), retrospective questionnaire provided at the time of delivery, with a spoken description of the content of the questionnaire. This poses, at least, two factors not adequately controlled. First, the actual frequency in which mothers spoke any of the two languages, as we rely only on their reports referring to the last trimester of pregnancy. Furthermore, although a minimum period of usage time had to occur to be considered as valid, the questionnaire did not address the exact amount of language usage within a day. Future studies should address these limitations, for instance, by collecting large amounts of data from a maternal diary of language usage during the last trimester of pregnancy and include an additional language abilities test (such as LEAP-Q; Marian et al., 2007) to evaluate the putative link between F₀ encoding abilities in newborns and maternal language usage percentage.

Overall, our findings emphasize the potential importance of prenatal linguistic exposure in shaping the neural mechanisms underlying language acquisition and highlight the sensitivity of the FFR in capturing these subtle changes. The results add to a growing body of research that suggests a role for prenatal fetal experiences in modeling language acquisition (Moon et al., 2012; Partanen et al., 2013b; Gervain, 2015, 2018; Arenillas-Alcón et al., 2023). Furthermore, they also highlight the importance of considering prenatal language exposure in developmental studies about language acquisition, a factor that is not routinely measured and reported, and that may contribute to divergent findings.

The present study contributes significant insights into the impact of prenatal bilingual exposure on the neural encoding of speech sounds at birth, thereby increasing our knowledge of the early stages of language acquisition. The observed differences in the encoding of voice pitch and formant structure depending on prenatal linguistic exposure highlight the remarkable plasticity and learning potential of the human brain even before birth, emphasizing the complex interaction between genetic and environmental factors in shaping our cognitive abilities and linguistic development.

The data supporting the conclusions of this article will be made available upon request by the authors, without undue reservation.

The studies involving humans were approved by Ethical Committee of Clinical Research of the Sant Joan de Déu Foundation (Approval ID: PIC-53-17). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants’ legal guardians/next of kin.

NG-C: Writing – review & editing, Writing – original draft, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. SA-A: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. MP: Investigation, Writing – review & editing, Methodology. AM-S: Writing – review & editing, Methodology, Investigation. SI-K: Writing – review & editing, Methodology, Investigation. JC-F: Writing – review & editing, Supervision, Methodology, Conceptualization. MG-R: Writing – review & editing, Resources, Funding acquisition. CE: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition, Conceptualization.